[
https://issues.apache.org/jira/browse/ARROW-16234?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17526658#comment-17526658
]
Will Ayd commented on ARROW-16234:
----------------------------------
I pushed up a rough draft for this on GH just to make sure the foundation was
right. However, I'm wondering if you think we should mirror what pandas does in
cases of ties or pick another default. Pandas interpolates an average for tied
rankings by default, which of course is going to change our returned data type.
Not sure if we want to stray from the integral return value as a default or
instead pick another thing like dense ranking
> [C++] Implement Rank Kernel
> ---------------------------
>
> Key: ARROW-16234
> URL: https://issues.apache.org/jira/browse/ARROW-16234
> Project: Apache Arrow
> Issue Type: Improvement
> Components: C++
> Reporter: Will Ayd
> Assignee: Will Ayd
> Priority: Minor
> Labels: C++, good-second-issue, kernel, pull-request-available
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Didn't see this in the library already so apologies if overlooked, but I
> think it would be nice to add a compute kernel for ranking. Here is a similar
> function in pandas:
> [https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.rank.html]
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
